The US General Services Administration (GSA), the federal agency responsible for managing the US government’s policies relating to procurement, property, and technology, has partnered with Meta to make the company’s open-source Llama artificial intelligence (AI) models more easily available to US federal agencies under its OneGov initiative.
The move is aimed at cutting red tape, reducing duplication, and speeding up the adoption of AI tools across government departments.
US Federal Acquisition Service Commissioner Josh Gruenbaum said that, in line with US President Donald Trump’s AI Action Plan, GSA is working to integrate AI into government operations.
He claimed that the OneGov initiative makes Llama and other open-source AI models more accessible to federal agencies, helping streamline processes and improve efficiency across departments.
When asked about deep discounts offered by tech firms to secure approval, Gruenbaum told Reuters, “It’s not about currying favour. It’s about recognising how we can all work together to make this country the best it can possibly be.”
Elsewhere, Meta CEO Mark Zuckerberg added that making Llama available to US government agencies could support more efficient public services and enable agencies to experiment with AI tools in practical applications.
Meta is not only working with US federal agencies but has also extended Llama access to key US allies. In consultation with the US government, Meta made its models available for defence and national security purposes to the US’ Five Eyes partners: Australia, Canada, New Zealand, and the UK.
The company is also now expanding access to France, Germany, Italy, Japan, South Korea, as well as North Atlantic Treaty Organisation (NATO) and European Union (EU) institutions.
Governments can run the models locally, allowing them to use the AI without sending sensitive information through third-party providers. Meta said this makes Llama suitable for classified or mission-specific work. US defence agencies have already piloted projects using Llama, including with the US Army’s Combined Arms Support Command to speed up equipment repairs using AI and virtual reality.
Meta is working with several private sector companies, such as Lockheed Martin, IBM, Microsoft, Palantir, Amazon Web Services, and Google Cloud, to implement Llama-based solutions for government and defence projects.
While Meta highlights the benefits of open-source AI, expanding into the defence and intelligence sectors raises broader questions. Open-source models are accessible to all, making them flexible and cost-effective, but also harder to control if misused. Providing wide access to AI in sensitive areas like national security requires strong oversight and safeguards.
The US government has already underlined the need for responsible use of AI in defence through the Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy. Meta has endorsed this principle, saying that countries must ensure AI is deployed “ethically, responsibly, and in accordance with relevant international law”.
Meta’s Llama is gaining traction in the US and its allied nations, just as Chinese rival DeepSeek faces mounting bans over security and privacy concerns.
The US Commerce Department has banned DeepSeek from government devices, with members of the US Congress and the country’s Navy flagging national security risks. And at least 14 US states, including Texas, New York, Virginia, and Georgia, have also issued statewide bans citing surveillance fears.
Globally, Italy blocked DeepSeek for failing to disclose data practices, and Taiwan, South Korea banned it within government departments and ministries, and Australia ordered its removal from all official government devices.
This split underscores a widening AI divide: while the US and its allies increasingly adopt Western models for governance, Chinese systems face growing distrust.
The US GSA has recently approved AI tools from Amazon Web Services, Microsoft, Google, Anthropic, and OpenAI, all of which agreed to sell their paid products to federal agencies at reduced prices after meeting security requirements.
Importantly, the GSA-Meta partnership highlights how AI is moving from experimental tools to core government operations, promising faster decision-making, reduced bureaucracy, and improved public services.
Also, expanding Llama access to US allies strengthens geopolitical ties and underscores the strategic importance of AI in defence and national security. But at the same time, open-source models raise concerns around oversight, security, and ethical deployment, making responsible governance critical.
